IOS Press Taxonomic and thematic organisation of proper

Behavioural Neurology 24 (2011) 265–276 265DOI 10.3233/BEN-2011-0337IOS Press

Taxonomic and thematic organisation ofproper name conceptual knowledge

Sebastian J. Crutch∗ and Elizabeth K. WarringtonDementia Research Centre, Department of Neurodegeneration, UCL Institute of Neurology, University CollegeLondon, London, UK

Abstract. We report the investigation of the organisation of proper names in two aphasic patients (NBC and FBI). The performanceof both patients on spoken word to written word matching tasks was inconsistent, affected by presentation rate and semanticrelatedness of the competing responses, all hallmarks of a refractory semantic access dysphasia. In a series of experiments weexplored the semantic relatedness effects within their proper name vocabulary, including brand names and person names. Firstwe demonstrated the interaction between very fine grain organisation and personal experience, with one patient with a specialinterest in the cinema demonstrating higher error rates when identifying the names of actors working in a similar film genre(e.g. action movies: Arnold Schwarzenegger, Bruce Willis, Sylvester Stallone, Mel Gibson) than those working in differentgenres (e.g. Arnold Schwarzenegger, Gregory Peck, Robin Williams, Gene Kelly). Second we compared directly two potentialprinciples of semantic organisation – taxonomic and thematic. Furthermore we considered these principles of organisation inthe context of the individuals’ personal knowledge base. We selected topics matching the interests and experience of eachpatient, namely cinema and literature (NBC) and naval history (FBI). The stimulus items were arranged in taxonomic arrays (e.g.Jane Austen, Emily Bronte, Agatha Christie), thematic arrays (e.g. Jane Austen, Pride and Prejudice, Mr Darcy), and unrelatedarrays (e.g. Jane Austen, Wuthering Heights, Hercule Poirot). We documented that different patterns of taxonomic and thematicorganisation were constrained by whether the individual has limited knowledge, moderate knowledge or detailed knowledgeof a particular vocabulary. It is suggested that moderate proper name knowledge is primarily organised by taxonomy whereasextensive experience results in a more detailed knowledge base in which theme is a powerful organising principle.

Keywords: Semantics, conceptual knowledge, proper noun, refractoriness

1. Introduction

Semantic refractory access dysphasia is a form ofcomprehension impairment in which access to storedconceptual knowledge is compromised. This type ofaccess disorder is typically characterised by responseinconsistency, sensitivity to temporal factors (presen-tation rate), sensitivity to semantic relatedness, and in-sensitivity to item frequency. Refractoriness is consid-ered to be a form of neural interference which leads to“reduction in the ability to utilise the system for a peri-od of time following activation” (Warrington and Mc-

∗Address for correspondence: Dr. Sebastian Crutch, DementiaResearch Centre, Box 16, National Hospital for Neurology and Neu-rosurgery, Queen Square, London, WC1N 3BG, UK. Tel.: +44 8451555 000 ext. 723113; Fax: +44 20 7676 2066; E-mail: [email protected].

Carthy, 1983, p. 874). For the purposes of the currentpaper, the critical aspect of the disorder is the sensitiv-ity to semantic relatedness. In semantic refractory ac-cess patients not only do individual, repeatedly probeditems become inaccessible but also semantically relat-ed items. Thus response accuracy is significantly lowerwhen identifying targets which are semantically relat-ed (e.g. furniture: chair, table, stool, desk, sofa, tele-vision) than semantically unrelated (e.g. chair, pencil,door, hat, foot, telephone).

The original investigations of refractoriness wereconducted using spoken word to picture matching tasksin the domain of common nouns. The build up of in-terference in these repetitive comprehension tasks wasshown to be caused by the semantic relatedness butnot the phonological relatedness of the items in the re-sponse array. For example, when provided with thespoken name ‘goat’, refractoriness can lead to greater

ISSN 0953-4180/11/$27.50 2011 – IOS Press and the authors. All rights reserved

266 S.J. Crutch and E.K. Warrington / Taxonomic and thematic organisation of proper name conceptual knowledge

difficulty pointing to a picture of a goat presented withpictures of other semantically related items (e.g. goat,deer, bear) than with other phonologically related items(e.g. goat, boat, coat [6,20]). Refractory effects havealso been explored using spoken word to written wordsmatching tests. This technique can be used to mea-sure semantic relatedness in patients with impairedphonological-orthographic transcoding. These match-ing procedures have been employed to investigate se-mantic relatedness in an abstract word vocabulary [7,9] and in a proper noun vocabulary [5,6,9]. For ex-ample, it has been shown that both brand names andperson names can become refractory when probed re-peatedly (e.g. higher error rates for the target ‘Ford’when presented in written arrays of semantically re-lated [i.e. Ford, Vauxhall, Rover, Honda] than unrelat-ed items [Ford, Tesco, Cadburys, Hitachi]). Measure-ments of semantic relatedness effect sizes can be takenas a marker of the similarity of a set of items, and thusbe used to assess the validity of hypothetical semanticdistinctions. Consequently one of the main motivationsfor studying patients with semantic refractory accessdysphasia is that they offer a window on the fine-grainorganisation of conceptual knowledge [19].

In our previous investigations of semantic refractoryaccess disorders, we have focussed upon two issues:(a) Evidence for fine-grain organisation within the se-mantic system, and (b) The different principles whichunderpin this organisation. One area of enquiry whichunites these two issues is the investigation of geograph-ical names [5,9]. It has been shown that refractory pa-tients find it more difficult to comprehend the namesof geographical terms (e.g. countries, cities, counties)when selecting a response from an array of geograph-ically close rather than geographically distant places.The influence of geographical location and proximityrevealed evidence of very fine grain organisation in thepatients’ knowledge of geographical names; geograph-ical proximity effects persisted even amongst Englishcities and London suburbs. Furthermore, these prox-imity effects were shown to cross taxonomic bound-aries. For example, geographically close town, regionand river names (e.g. London, Surrey, Thames) elicitedhigher error rates that names drawn from the same taxo-nomic class (e.g. Thames, Loire, Rhine). These Exper-iments suggest that at least two principles (taxonomicand geographic proximity) govern the organisation ofour understanding of geographical names.

In this paper we explore a principle other than tax-onomy which guides the organisation of other classesof proper names. For the purpose of these experiments,

this additional principle of organisation is broadly la-belled ‘thematic’. This distinction between taxonomyand theme has been anticipated in a number of previousstudies [10,15,16] and a number of terms have beenused to capture such different forms of semantic rela-tionship (e.g. similarity-contiguity, paradigm-syntagm;Coltheart, 1981). It is well recognised that our concep-tual system allows for the representation of differentforms of semantic relationship such as category mem-bership [1], and associative connections based on sub-ject or expert report [12,14]. Indeed it has also beensuggested that concrete and abstract words may differin their dependence upon similarity-based and associa-tive information [7]. Barsalou and colleagues suggestthat information is clustered more by theme than tax-onomy for abstract words (and vice versa for concreteconcepts) and argue that thematic and taxonomic clus-tering reflects a drive to collect concepts into groupswhich support goal-directed action [22].

A third issue central to our investigation of seman-tic refractory access disorders is the influence of per-sonal experience upon the organisation of conceptualknowledge. In the geographical experiments describedabove, the probing of geographical names was predi-cated upon the assumption that patients had knowledgeof their position within a spatial framework. This as-sumption was tested by comparing their comprehen-sion of domestic and foreign place names. For exam-ple, for the English patients tested, geographical relat-edness effects were observed among UK city namesbut not among US state names. We have argued thatpersonal knowledge and experience interact with thesemantic relatedness effects observed in this class ofpatient.

In the current study, our aim was to examine whetherdifferent principles of semantic organisation could bedirectly compared and contrasted in a variety of propernoun vocabularies. It was hypothesised that, within therealms of personal knowledge, more powerful forms ofsemantic relatedness than mere taxonomic class mightemerge. This hypothesis was tested by developingstimulus sets which corresponded to the specific inter-ests and experience of a particular patient (e.g. cine-ma, naval history). In the patient into whose particularinterests the stimuli were designed to tap, it was an-ticipated that thematic relatedness would elicit greatersemantic interference (as measured by high error rate)than mere taxonomic relatedness. By contrast, in thepatient who did not share that particular interest, it waspredicted that semantic interference would be limitedto taxonomically related stimuli or not occur at all.

S.J. Crutch and E.K. Warrington / Taxonomic and thematic organisation of proper name conceptual knowledge 267

Table 1The number and percentage correct responses on the background screening tests

NBC FBI

WASI Matrix Reasoning 9/32 (T = 34) 10/32 (T = 38)Camden Pictorial Memory Test 14/30 (< 5th %ile) 30/30 (> 10th %ile)Digit span 0 digits forwards 2 digits forwardsRepetition 0/90 61/90Reading (from Brown and Ure, 1969) 0/72 25/72Graded Nonword Reading Test 0/20 0/20Spoken nonword-written nonword match 11/20 3/10British Picture Vocabulary Scale (short) 10/32 20/32McKenna and Warrington (1978)Naming Animals 2/10 (20%) 6/10 (60%)

Man-made artefacts 0/10 (0%) 2/10 (20%)Colours 1/10 (10%) 5/10 (50%)Body part 1/10 (10%) 1/10 (10%)Countries 1/10 (10%) 7/10 (70%)TOTAL 5/50 (10%) 21/50 (42%)

Comprehension Animals 8/10 (80%) 10/10 (100%)Man-made artefacts 10/10 (100%) 7/10 (70%)Colours 6/10 (60%) 7/10 (70%)Body part 2/10 (20%) 1/10 (10%)Countries 10/10 (100%) 10/10 (100%)TOTAL 36/50 (72%) 35/50 (70%)

Semantic refractory access dysphasia screening testsPresentation rate Fast (1s RSI) 103/128 (80%) 104/128 (81%)

Slow (10s RSI) 117/128 (91%) 121/128 (95%)Semantic relatedness Related 100/128 (78%) 106/128 (83%)

Unrelated 120/128 (94%) 119/128 (93%)Frequency Low frequency 20/32 (63%) 29/32 (91%)

Middle frequency 21/32 (66%) 27/32 (84%)High frequency 22/32 (69%) 30/32 (94%)

2. Participants

NBC is a 64-year-old retired accountantwho becameglobally aphasic following a left middle cerebral arterystroke in 2007. MRI confirmed a left MCA infarct anda left cerebellar infarct. This patient has been report-ed in a two previous studies (Crutch and Warrington,2010; Crutch and Jackson, in press). NBC exhibiteda very keen interest in film and cinema, and contin-ued to attend film showings and peruse magazines andnewspaper articles about famous actors and actresses.

FBI is a 70-year-old, right-handed retired sailor andbuilding maintenance engineer who became globallyaphasic following a large left middle cerebral arteryischaemic stroke in March 2005. A computed tomog-raphy (CT) scan conducted on the day of admissionrevealed a left parietal infarct with minimal oedemaand mass effect. No further imaging is available. In-formation about this patient may be found in severalprevious single case and case series reports [8,9]. FBIwas severely physically disabled by his stroke but ex-pressed particular enthusiasm for documentaries andbooks about naval matters.

These patients were included in this study becausethey both showed (a) the characteristic features of se-

mantic refractory access dysphasia, and (b) impairedphonological-orthographic transcoding so that spokenword towrittenwordmatching tasks could be employedas a measure of semantic processing (see below forconfirmatory details of the patient’s neuropsychologi-cal profiles).

3. Background information

NBC completed a short battery of tests of neuropsy-chological function (see Table 1). He showed lowaverage-borderline non-verbal reasoning performanceon the Matrix Reasoning subtest from the WechslerAbbreviated Scale of Intelligence [21]. However heappeared unable to understand the Camden PictorialMemory Test [18]. Phonological deficits were appar-ent in his failure to read any items from the GradedNonword Reading test [17], and his chance score whenrequired to point to a specified cvc given a choice of twononwords with no letters in common (e.g. fip – dag).He was also completely unable to repeat or read a singleword aloud including single digits. NBC demonstrateda severe anomia, managing to produce the names of


very few of the high frequency items from McKennaand Warrington [13]. He obtained a weak score on theshort form of the British Picture Vocabulary Scale [11],and showed impaired performance on a simple spokenword-picture test involving high frequency items [4,13]. In a series of spoken word to picture matchingtasks, NBC’s comprehension performance was shownto be (a) sensitive to presentation rate, (b) sensitive tosemantic relatedness, but (c) insensitive to item fre-quency (see Table 1).

FBI completed the same short battery of tests of neu-ropsychological function as NBC (see Table 1). Heshowed low average-borderline non-verbal reasoningperformance on the Matrix Reasoning subtest from theWASI. Visual episodic memory was well preserved asindicated by strong performance on the Camden Pic-torial Memory Test. However, short term memory asmeasured by digit span was severely impaired. FBIwas able to read and repeat a large proportion of thesingle words presented (see Warrington and Crutch,2007, for full details of FBI’s dyslexia). Phonologicalskills were severely compromised as he failed to readany items from the Graded Nonword Reading test, andscored at chance when required to point to a specifiedcvc given a choice of two nonwords with no letters incommon (e.g. fip – dag). Naming was also severelycompromised, with FBI producing the names of lessthan half of the high frequency items from McKennaand Warrington [13]. On tests of semantic processing,he obtained a weak score on the BPVS, and this com-prehension deficit was mirrored by impaired perfor-mance on the simple spoken word-picture tests involv-ing high frequency items. On spoken word to picturematching tasks, FBI was also shown to be sensitive topresentation rate and semantic relatedness but not itemfrequency (see Table 1).

The neuropsychological profiles indicate that bothpatients exhibited global aphasia. The combinationof the hallmark features of semantic refractory accessdyslexia and these patients’ impaired phonological-orthographic transcoding meant that it was thereforeappropriate to administer repetitive spoken word towritten word matching tasks as a measure of semanticprocessing.

4. Experimental investigation

All of the experiments reported in this investigationemploy a spoken word to written word matching pro-cedure, in which the patients were presented with an

array of 3 or 4 written words and then the examinerspoke the name of each word aloud several times ina fixed random order, on each occasion requiring thepatient to point to the word they had just heard. Asnoted above, this procedure was an appropriate meansof investigating the patients’ comprehension of thesewords because their abolished nonword reading andspoken-nonword-to-written-nonwordmatching perfor-mance indicated minimal residual capacity for phono-logical processing of the written word. Thus any ac-curate word reading and spoken-written matching washeld to reflect semantic processing.

Experiment 1 – Semantic relatedness effects in brandnames and person names

Before comparing directly the influence of taxonom-ic and thematic relatedness in multiple proper noun do-mains, we examined whether the semantic relatednesseffects observed previously in NBC and FBI’s compre-hension of place names were also observed in the prop-er noun domains of brand names and people. Thesetests constitute a replication (Experiment 1a) and ex-tension (Experiment 1b) of previous single case evi-dence of proper noun relatedness effects [6]. In partic-ular, patient NBC had reported a particular interest incinema, so a novel task involving famous film actorswas administered.

MethodsThe comprehension of two sets of proper nouns were

examined:

a) Brand names – these test materials were previ-ously described in Crutch and Warrington (2004;Experiment 3). The stimuli consisted of 16 well-known brand names, four drawn from each of thecategories of supermarkets, car manufacturers,food producers and electrical goods manufactur-ers. These were arranged into four semanticallyclose arrays (e.g. Safeway, Tesco, Waitrose, andAsda) and four semantically distant arrays (e.g.Safeway, Rover, Kelloggs, and Panasonic).

b) Person names (cinema) – the names of 16 fa-mous actors were probed under two relatednessconditions. Semantically related arrays involvedthe names of four actors who work primarily ina similar cinematic genre (e.g. action movies:Arnold Schwarzenegger, Bruce Willis, SylvesterStallone, Mel Gibson). Semantically unrelatedarrays involved the same names rearranged so that


each array included one actor from each differ-ent genre (e.g. Arnold Schwarzenegger, GregoryPeck, Robin Williams, Gene Kelly).

All stimuli are shown in the Appendix. The two sets ofstimuli were administered on separate occasions, andall arrays were presented in an ABBA design with anapproximate 1s responses stimulus interval.

ResultsThe number and percentage correct responses on

each task in Experiment 1 are shown in Table 2. Per-formance in the related and unrelated conditions wasevaluated using Wilcoxon matched-pairs signed-rankstests. On the brand names task, both patients showedsignificant semantic relatedness effects, respondingless accuratelywith related than unrelated brand names.On the specialised person task involving film actors,only NBC who was particularly interested in cinemashowed a significant relatedness effect.

CommentThese results suggest that the comprehension perfor-

mance of both patients on spoken word to written wordmatching tasks involving proper noun domains otherthan geographical terms was sensitive to semantic re-latedness. More specifically, as is typical of patientswith semantic refractory access dysphasia, response ac-curacy was reduced in conditions of semantic related-ness, indicating greater semantic interference amongthe item representations. These data replicate and ex-tend those reported previously for another semantic re-fractory access dysphasic patient, AZ [6].

Experiment 2 – Direct comparisons of taxonomic andthematic relatedness

In the second section of the investigation, semanticrelatedness effects were exploited to directly compareand contrast the influence of two different types of in-formation upon item comprehension: taxonomic classand thematic association1. As noted in the Introduc-tion, sensitivity to semantic relatedness is constrainedby the amount of knowledge available (e.g. through per-sonal experience, education). Here we explore the im-pact of personal knowledge upon word comprehensionacross several domains of conceptual knowledge.

1It should be noted that taxonomic class and thematic associa-tion are not mutually exclusive, as items within a taxonomic classoften co-occur in our linguistic or real world experience. By ref-erence to other terms used in the aphasic literature, the essentialdistinction of interest here is between similarity/shared features andco-occurrence/association.

MethodsThe influence of taxonomic and thematic relatedness

upon comprehension performance (relative to an unre-lated control condition)was examined in 4 different vo-cabularies. Two of these domains of knowledge wereselected to match NBC’s main interests (literature andfilm), and 2 were selected to match FBI’s main interests(naval history; see Appendix). In all cases, the overallexperimental design was identical. Stimuli (N = 9or 16) were tested under three conditions – taxonom-ic, thematic, and unrelated – with conditions varied inan ABCABC or Latin square design. The domains ofknowledge examined were:

a) Literature – the stimuli were authors, novels andcharacters, arranged as follows:

– Taxonomy: 3 authors (e.g. Jane Austen, EmilyBronte, Agatha Christie), 3 novels and 3 char-acters

– Thematic: Related author-novel-charactergro-upings (e.g. Jane Austen, Pride and Prejudice,Mr Darcy)

– Unrelated: Unrelated author-novel-charactergroupings (e.g. Jane Austen, Wuthering Heig-hts, Hercule Poirot)

b) Film – the stimuli were the names of films andtheir respective directors, actors and actresses,arranged as follows:

– Taxonomy: 4 films (e.g. Chinatown,Rear Win-dow, Taxi Driver, Some Like it Hot), 4 direc-tors, 4 actors and 4 actresses

– Thematic: Related film-director-actor-actressgroupings (e.g. Chinatown, Roman Polanski,Jack Nicholson, Faye Dunaway)

– Unrelated: Unrelated film-director-actor-actr-ess groupings (e.g. Chinatown, Billy Wilder,Robert De Niro, Grace Kelly)

c) Naval explorers – the stimuli were the names fa-mous explorers, their corresponding vessels, andthe locationswith which they are most associated,arranged as follows:

– Taxonomy: 3 explorers (e.g. Ernest Shackle-ton, Christopher Columbus, Captain Bligh), 3vessels and 3 locations

– Thematic: Related explorer-vessel-location gr-oupings (e.g. Ernest Shackleton, Endurance,Antarctica)

– Unrelated: Unrelated explorer-vessel-locationgroupings (e.g. Ernest Shackleton, Santa Mar-ia, Tahiti)


Table 2Number and percentage correct responses in each related and unrelatedcondition in Experiment 1

Related Unrelated Related vs unrelated

1a Brands NBC 34/64 59/64 3.19(53%) (92%) P = 0.001

FBI 33/64 57/64 3.41(52%) (89%) P = 0.0007

1b Actors NBC 47/64 58/64 2.04(73%) (91%) P = 0.04

FBI 37/64 45/64 0.89(58%) (70%) P > 0.3

Table 3Number and percentage correct responses in each taxonomic, thematic and unrelated condition in Experiment 2.Wilcoxon matched-pairs signed-ranks statistical pairwise comparisons of taxonomic, thematic and unrelated conditions

Taxonomic Thematic Unrelated Taxonomic vs Taxonomic vs Thematic vsthematic unrelated unrelated

2a Literature NBC 35/36 26/36 35/36 −2.14 0.00 2.39(97%) (72%) (97%) P = 0.03 P = 1 P = 0.02

FBI 23/36 35/36 30/36 2.19 1.51 −1.73(64%) (97%) (83%) P = 0.03 P = 0.13 P = 0.08

2b Film NBC 56/64 42/64 58/64 −2.36 0.18 3.09(88%) (66%) (91%) P = 0.02 P = 0.86 P = 0.002

FBI 46/64 57/64 54/64 2.22 1.52 −1.00(72%) (89%) (84%) P = 0.03 P = 0.13 P = 0.32

2c Explorers NBC 32/36 21/36 29/36 −2.21 −1.34 1.81(89%) (58%) (81%) P = 0.03 P = 0.18 P = 0.07

FBI 36/36 29/36 35/36 −2.41 −1.00 1.91(100%) (81%) (97%) P = 0.02 P = 0.32 P = 0.06

2d Naval NBC 28/36 28/36 33/36 −0.06 0.93 1.04battles (78%) (78%) (92%) P = 0.95 P = 0.35 P = 0.30

FBI 18/36 33/36 28/36 2.27 2.39 −1.45(50%) (92%) (78%) P = 0.02 P = 0.02 P = 0.15

d) Naval battles – the stimuli were the names ofrenowned naval battles and the commanders andvessels most associated with them, arranged asfollows:

– Taxonomy: 3 naval battles (e.g. Jutland, Mat-apan, Falklands), 3 commanders and 3 vessels

– Thematic: Related battle-commander-vesselgr- oupings (e.g. Jutland, Jellicoe, Indefa-tiguable)

– Unrelated: Unrelated battle-commander-vess-el groupings (e.g. Jutland, Cunningham, Bel-grano)

ResultsThe number and percentage correct response on Ex-

periments 2a-d are shown in Table 3 and Fig. 1. In con-sidering the results, it should be noted that, paradoxical-ly, higher error rates are considered to indicate a higherdegree of knowledge of the stimuli being tested (andhence a greater susceptibility to the build up of semanticinterference). The corollary is that lower error rates are

considered to indicate a lesser degree of knowledge ofthe stimuli (and hence lower susceptibility to semanticinterference). The unrelated arrays condition providesan estimate of baseline word knowledge.

Experiments 2a - LiteratureOn the first set of stimuli selected to match NBC –

his interest in and knowledge of literature – NBC’sperformance was most impaired in the thematically re-lated condition as compared with both the taxonomicand unrelated conditions. For example, his error ratewas much higher when asked simply to point to a the-matically associated triplet of names (e.g. Jane Austen,Pride and Prejudice, Mr Darcy) than a taxonomicallyrelated triplet of the names of three authors (e.g. JaneAusten, Emily Bronte,Agatha Christie). This is despitethe many common features of the taxonomic items (e.g.gender, profession, etc.). This was also observed in thecontext of his baseline performance of 97% correct inthe unrelated condition.

By contrast, FBI showed a quite different pattern ofperformance across the three conditions, with the high-


Fig. 1. The percentage correct responses in Experiment 2.

est error rate in the taxonomically related condition.There was a trend toward the opposite pattern to NBC,namely a higher error rate on the taxonomically relatedas compared to thematically related condition.

Experiment 2b – FilmOn the second set of stimuli selected to suitNBC–his

keen interest in film and cinema – NBC’s performancewas again most impaired in the thematically relatedcondition. He showed significantly higher error rate inthe thematically related condition (e.g. Rear Window,Alfred Hitchcock, James Stewart, Grace Kelly) thaneither the taxonomically related (e.g. Roman Polans-ki, Alfred Hitchcock, Martin Scorsese, Billy Wilder)or unrelated conditions. However there were no com-parable differences between his performances on thetaxonomically related and unrelated conditions.

Patient FBI was apparently insensitive to these di-mensions, scoring at a satisfactory level on all threeconditions and with no significant differences betweeneither the thematic or taxonomic conditions relative tothe unrelated condition.

Experiments 2c – ExplorersThe third set of stimuli was selected to match FBI’s

areas of personal interest and knowledge in naval histo-ry. On the Explorers task (Experiment 2c), both NBCand FBI showed a mild thematic effect. Both showedstatistical trends toward lower scores on the thematiccondition (e.g. Christopher Columbus, America, San-ta Maria) than the control unrelated condition in bothpatients, but there were no differences between per-formances on the taxonomically related (e.g. Christo-pher Columbus, Ernest Shackleton, Captain Bligh) andunrelated conditions.

Experiment 2d – Naval battlesOn the more demanding Naval battles task, NBC

showed no pairwise differences between conditions.However, FBI showed significantly lower performanceon the taxonomic (e.g. Matapan, Jutland, Falklands)than unrelated condition. There was no accompanyingdifference between the thematic (e.g. Jellicoe, Jutland,Indefatiguable) and unrelated conditions.


Comment

Patient NBC showed significant thematic effects in3/4 Experiments, including both the tasks designed tofocus on his particular areas of knowledge and interest(literature and cinema) and also one of the tasks de-signed with patient FBI in mind (explorers). By con-trast patient FBI showed no thematic or taxonomic ef-fects on tasks examining literature or cinema, thoughby his adequate performance in the unrelated condi-tions demonstrated these items were clearly within hisvocabulary. However, FBI did show a thematic effectwith explorers, and perhaps more impressively a sig-nificant taxonomic effect with naval battles, indicatingfamiliarity with this obscure topic.

5. Discussion

In this study we have explored aspects of propername knowledge hitherto neglected. In particular, weexamined the principles and extent of fine-grain organ-isation within this conceptual domain, and also consid-ered the influence of personal experience upon com-prehension performance.

In Experiment 1, we reported data indicating thefine-grain organisation of both brand names and personnames in two semantic refractory access dysphasic pa-tients, NBC and FBI. For brand names, semantic refrac-toriness was greater (yielding higher error rates) for re-lated (e.g. Safeway, Tesco, Waitrose, Asda) than unre-lated (e.g. Safeway, Rover, Kelloggs, Panasonic) items.This performance replicated that of patient AZ [6].The impact of personal experience upon the extent offine grain organisation was then explored using stimulidrawn from the area of expertise of one patient (NBC)but not the other (FBI), namely male film actors asso-ciated with a particular cinematic genre. Patient NBCshowed significantly higher errors rates when identify-ing the names of actors who worked in a similar genre(e.g. action movies: Arnold Schwarzenegger, BruceWillis, Sylvester Stallone, Mel Gibson) than those whoworked in different genres (e.g. Arnold Schwarzeneg-ger, Gregory Peck, Robin Williams, Gene Kelly). Pa-tient FBI was relatively insensitive to this manipu-lation. This pattern of refractory performance indi-cates fine grain organisation of person knowledge notjust between occupations (as previously documented inCrutch and Warrington [6]) but even within a singleoccupation. The caveat here is that the absolute level

of fine grain organisation is dictated to a large extentby personal experience and other individual factors.

In Experiment 2, we compared directly two poten-tial principles of semantic organisation – taxonomy andtheme – across domains of knowledge in which our twopatients had varied levels of expertise. In the areas ofliterature and cinema (Experiments 2a and 2b), NBCshowed greater refractoriness (shown by significantlyhigher error rates) in the thematic condition than in thetaxonomic and unrelated conditions. Of particular in-terest is the observation in Experiment 2a that associ-ated concepts (e.g. Jane Austen, Pride and Prejudice,Mr Darcy) generated more semantic interference thantaxonomically similar concepts, despite the fact thatthe latter share many more semantic features (e.g. JaneAusten, Emily Bronte, Agatha Christie – all female, allnovelists, all deceased, etc.). By contrast, FBI showedhigher error rates in the taxonomic condition, suggest-ing that he had moderate but not as detailed knowledgeas NBC about these authors, novels and characters.This suggests that within NBC’s conceptual knowledgebase, thematically-related items (e.g. authors, novels,characters) are represented more closely in semanticspace than taxonomically-related items (e.g. individu-als from the category of authors).

In areas of particular personal knowledge (naval his-tory) for FBI, both patients in fact showed significantlymore semantic interference in the thematic condition ofthe Explorers task (Experiment 2c). It should be not-ed thought that the stories of all three of the explorersselected are relatively well-known and have been madeinto mainstream films, which might account for whynot only FBI’s but also NBC’s performance indicatedfamiliarity with the specific associations between dif-ferent explorers, ships and destinations. On the moredemanding and obscure naval battles test however, itwas only FBI who showed any significant differencesin performancebetween conditions. He showed greatersemantic interference in the taxonomic than thematicor unrelated conditions, suggesting he possessed strongknowledge for the battle, ship and commander namesused but perhaps did not retain adequate knowledgeto be clear on the exact thematic connections. NBCshowed no significant evidence of semantic relatednesseffects on this task, suggesting weak knowledge of thistopic.

Three broad patterns of performance emerged fromExperiment 2. For the purposes of this Discussionwe label these patterns ‘limited knowledge’, ‘moderateknowledge’ and ‘detailed knowledge’, and these are il-lustrated schematically in Fig. 2. ‘Limited knowledge’,


Fig. 2. Hypothetical patterns of performance across the taxonomic, thematic and unrelated conditions for different degrees of personal knowledge.

as demonstrated by NBC on task 2d, indicates that theverbal labels are within one’s vocabulary but little moredetail is available (e.g. familiar with the name Jellicoebut no idea what he does). As a consequence, no sig-nificant difference is observed between related and un-related arrays. ‘Moderate knowledge’, as demonstrat-ed by FBI on tasks 2a, 2b and 2d, indicates enoughknowledge has been acquired to achieve a taxonomicclassification of the term but insufficient detail to haveindividual thematic associations (e.g. know that MartinScorsese is afilm director but notwhich individualfilmshe is associated with). As a consequence, taxonomicarrays yield a significantly higher error rate than the-matic and/or unrelated arrays. ‘Detailed knowledge’,as demonstrated by NBC on tasks 2a-c and FBI ontask 2c, indicates that individual thematic associationsbetween concepts have been acquired (e.g. know thatMartin Scorsese is a film director and that he directedTaxi Driver). In this instance, a v-shaped pattern isobserved in which thematic arrays yield a significantlyhigher error rate than taxonomic or unrelated arrays.

The counterintuitive aspect of this pattern of perfor-mance is that there may be no significant difference be-tween taxonomically related and unrelated items. Thisabsence of a taxonomic effect does not mean that infor-mation about taxonomic class has been lost, but ratherthat the taxonomic effect has been over-taken by a morepowerful thematic effect. An analogy could be drawnhere to the organisationwithin small and large libraries:

in a very small library, Agatha Christie and Jane Austennovels might both be found on the ‘Fiction’ bookcase,whereas in a larger library they might be found on sep-arate bookcases for ‘Crime’ and ‘Classics’. Thus amore detailed classification can increase the distancebetween items which are taxonomically related. In thehuman brain, this organisation of psychological spaceis mirrored in neural organisation, with the representa-tions or patterns of activation underpinning taxonomi-cally related concepts becoming more distinct with in-creasing knowledge. This increase in taxonomic se-mantic distance with increasing experience is a fea-ture of the computational models of Rogers and Mc-Clelland (2004). By training networks to be ‘experts’in either bird or fish knowledge by manipulating theexposure during the training phase, these researchersdemonstrated a tendency to name at the subordinatelevel in the expert domain and at the basic level in thenon-expert domain. The greater differentiation of sub-ordinate items in the expert domain was also demon-strated by greater mean Euclidean distances betweenthe networks internal representations.

A few further points regarding patient performanceand experimental design are worth noting. First,the experimental approach described is dependent onpatients having sufficiently impaired phonological-orthographic transcoding as to permit spoken word towritten word matching tasks to be used as a mea-sure of semantic processing. Nonetheless both patients


demonstrated a remarkable insensitivity to variablessuch as the number of words in the stimulus (e.g. point-ing to ‘Death on the Nile’ in response to the auditoryprobe ‘Agatha Christie’). Second, one potential limita-tion to this line of investigation is that the experimentswere tailor-made to the patients we were studying. Notall individuals have an area of specialist knowledge orinterest, and it is not possible to match or balance forthe extent of this knowledge across subject areas. Nev-ertheless, it can be argued that single case methodologyprovides an opportunity to study aspects of the cerebralorganisation of personal knowledge which would bedifficult within a group study design.

In conclusion, many investigations of conceptualknowledge have examined a vocabulary which is fa-miliar to the majority of individuals within a society.However, there are many aspects of personal experi-ence and education which have a powerful influence onhow this common database is extended. The currentstudy points to the role of both taxonomic and themat-ic information in the fine grain organisation of prop-er name knowledge. These principles play a criticalrole in creating the personalised cerebral knowledgebase through which we understand and interact withthe world.

Acknowledgements

We would like to thank Dr David Cohen and the ther-apists on the Harrow Stroke Unit and the Regional Re-habilitation Unit at Northwick Park Hospital for allow-ing us to work with patients under their care. This workwas undertaken at UCLH/UCL who received a propor-tion of funding from the Department of Health’s NIHRBiomedical Research Centres funding scheme. TheDementia Research Centre is an Alzheimer’s ResearchTrust Co-ordinating Centre. This work was support-ed by an Alzheimer’s Research Trust Senior ResearchFellowship to SC.

References

[1] J.R. Anderson, The adaptive nature of human categorization,Psychological Review 98 (1991), 409–429.

[2] M. Coltheart, The semantic error: Types and theories, in:Deep Dyslexia, M. Coltheart, K. Patterson and J.C. Marshall,eds, London: Routledge and Kegan Paul, 1980, 146–159.

[3] S.J. Crutch and E.C. Jackson, Contrasting graded effects ofsemantic similarity and association across the concreteness

spectrum. Quarterly Journal of Experimental Psychology (inpress).

[4] S.J. Crutch, K. Randlesome and E.K. Warrington, The vari-ability of country map knowledge in normal and aphasic sub-jects: evidence from two new category-specific screeningtests, J Neuropsychol 1 (2007), 171–187.

[5] S.J. Crutch and E.K. Warrington, The human brain map:knowledge of country and city names depends on their geo-graphical proximity, Brain 126 (2003), 1821–1829.

[6] S.J. Crutch and E.K. Warrington, The semantic organisationof proper nouns: the case of people and brand names, Neu-ropsychologia 42 (2004), 584–596.

[7] S.J. Crutch and E.K. Warrington, Abstract and concrete con-cepts have structurally different representational frameworks,Brain 128 (2005), 615–627.

[8] S.J. Crutch and E.K. Warrington, Contrasting patterns of com-prehension for general and specific semantic information insemantic dementia and stroke, Cognitive Neuropsychology 25(2008), 582–600.

[9] S.J. Crutch and E.K. Warrington, Spatially coded semanticinformation about geographical terms, Neuropsychologia 48(2010), 2120–2129.

[10] J. Davidoff and D. Roberson, Preserved thematic and impairedtaxonomic categorisation: a case study, Language and Cogni-tive Processes 19 (2004), 137–174.

[11] L.M. Dunn, L.M. Dunn, C. Whetton and D. Pintilie, TheBritish Picture Vocabulary Scale. Windsor, UK: NFER – Nel-son, (1982).

[12] M. Jarmasz and S. Szpakowicz, Roget’s thesaurus and seman-tic similarity, Proceedings of the International Conference onRecent Advances in Natural Language Processing (RANLP-03), Borovets, Bulgaria, September, 2003, pp. 212–219.

[13] P. McKenna and E.K. Warrington, Category-specific namingpreservation: a single case study, J Neurol Neurosurg Psychi-atry 41 (1978), 571–574.

[14] D.L. Nelson, C.L. McEvoy and T.A. Schreiber, The Universi-ty of South Florida word association, rhyme, and word frag-ment norms, 1998. Available from http://www.usf.edu/FreeAssociation/.

[15] C. Semenza, P. Bisiacchi and L. Romani, Naming disordersand semantic representations, Journal of Psycholinguistic Re-search 21 (1992), 349–364.

[16] C. Semenza, G. Denes, D. Lucchese and P. Bisiacchi, Selec-tive deficit of conceptual structures in aphasia: Class versusthematic relations, Brain and Language 10 (1980), 243–248.

[17] M. Snowling, S.E. Stothard and J. McLean, The Graded Non-word Reading Test. Bury St Edmunds: Thames Valley TestCompany, 1996.

[18] E.K. Warrington, The Camden Memory Test Manual. Hove,East Sussex: Psychology Press, 1996.

[19] E.K.Warrington andS.Crutch, Semantic refractory access dis-orders, in: The Neural Bases of Semantic Memory, J. Hart andM.A. Kraut, eds, Cambridge: Cambridge University Press,2007, pp. 3–27.

[20] E.K. Warrington and R. McCarthy, Category specific accessdysphasia, Brain 106 (1983), 859–878.

[21] D. Wechsler, The Wechsler Abbreviated Scale of Intelligence.St Antonio: The Psychological Corporation, 1999.

[22] C.D. Wilson, S.T. Gries, J.L. Brasfield and L.W. Barsalou, Hi-erarchical clustering of abstract and concrete concepts. Meet-ing of the Cognitive Science Society, Amsterdam, July 2009.


Appendix

Experiment 1bSemantically relatedScreen legends Action movie actors Comic actors Musical/Dance actorsGregory Peck Arnold

SchwarzeneggerRobin Williams Gene Kelly

James Stewart Bruce Willis Steve Martin Fred AstaireHumphrey Bogart Sylvester Stallone Jim Carrey Mickey RooneyClark Gable Mel Gibson Chevy Chase Frank SinatraSemantically unrelatedGregory Peck James Stewart Humphrey Bogart Clark GableArnold Schwarzenegger Bruce Willis Sylvester Stallone Mel GibsonRobin Williams Steve Martin Jim Carrey Chevy ChaseGene Kelly Fred Astaire Mickey Rooney Frank SinatraExperiment 2aTaxonomically relatedAuthors Novels CharactersJane Austen Pride and Prejudice Mr DarcyEmily Bronte Wuthering Heights HeathcliffAgatha Christie Death on the Nile Hercule PoirotThematically relatedJane Austen Emily Bronte Agatha ChristiePride and Prejudice Wuthering Heights Death on the NileMr Darcy Heathcliff Hercule PoirotUnrelatedJane Austen Death on the Nile Pride and PrejudiceWuthering Heights Emily Bronte HeathcliffHercule Poirot Mr Darcy Agatha ChristieExperiment 2bTaxonomically relatedFilms Directors Actors ActressChinatown Roman Polanski Jack Nicholson Faye DunawayRear Window Alfred Hitchcock James Stewart Grace KellyTaxi Driver Martin Scorsese Robert De Niro Jodie FosterSome Like it Hot Billy Wilder Jack Lemmon Marilyn MonroeThematically relatedChinatown Rear Window Taxi Driver Some Like it HotRoman Polanski Alfred Hitchcock Martin Scorsese Billy WilderJack Nicholson James Stewart Robert De Niro Jack LemmonFaye Dunaway Grace Kelly Jodie Foster Marilyn MonroeUnrelatedRear Window James Stewart Robert De Niro Jodie FosterMartin Scorsese Marilyn Monroe Grace Kelly Alfred HitchcockJack Lemmon Taxi Driver Chinatown Jack NicholsonFaye Dunaway Roman Polanski Billy Wilder Some Like it HotExperiment 2cTaxonomically relatedVessel Explorer/Captain LocationHMS Bounty Captain Bligh TahitiSanta Maria Christopher

ColumbusAmerica

Endurance Ernest Shackleton AntarcticaThematically relatedHMS Bounty Santa Maria EnduranceCaptain Bligh Christopher

ColumbusErnest Shackleton

Tahiti America Antarctica


UnrelatedSanta Maria Christopher

ColumbusAmerica

Ernest Shackleton Antarctica EnduranceTahiti HMS Bounty Captain BlighExperiment 2dTaxonomically relatedVessel Commander BattleWarsprite Cunningham MatapanIndefatiguable Jellicoe JutlandBelgrano Woodward FalklandsThematically relatedWarsprite Indefatiguable BelgranoCunningham Jellicoe WoodwardMatapan Jutland FalklandsUnrelatedBelgrano Indefatiguable FalklandsJutland Woodward WarspriteCunningham Matapan Jellicoe

Submit your manuscripts athttp://www.hindawi.com

Stem CellsInternational

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014


MEDIATORSINFLAMMATION

of


Behavioural Neurology

EndocrinologyInternational Journal of



Disease Markers


BioMed Research International

OncologyJournal of



Oxidative Medicine and Cellular Longevity


PPAR Research

The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014

Immunology ResearchHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Journal of

ObesityJournal of



Computational and Mathematical Methods in Medicine

OphthalmologyJournal of


Diabetes ResearchJournal of



Research and TreatmentAIDS


Gastroenterology Research and Practice


Parkinson’s Disease

Evidence-Based Complementary and Alternative Medicine

Volume 2014Hindawi Publishing Corporationhttp://www.hindawi.com

Documents

IOS Press Taxonomic and thematic organisation of proper